Several libraries already exist to provide a common API for all type of GNNs, like scikit-learn does for machine learning algorithms. In Python, you can refer to any of the following, depending on your favorite deep learning package:
- PyTorch Geometric: As the name suggests, this is a PyTorch extension that allows us to deal with complex datasets such as graphs with a new Dataset object. It also gathers tens of algorithm implementations (https://github.com/rusty1s/pytorch_geometric).
- Graph Nets library: Created by DeepMind, the company behind AlphaGo, the algorithm that was first that was able to beat a human player at Go. With Graph Nets, you will be able to build GNNs using TensorFlow (https://github.com/deepmind/graph_nets).
- Deep Graph Library (DGL): Supporting both PyTorch and TensorFlow, DGL provides tools to build all types of GNNs (https://www.dgl.ai/).
- GDS: Starting from its version 1.3, the GDS contains implementations for some embedding algorithms...
